Actionable AI: Data science meets social science

machine learning

Replaceable. How many of you hear that word in the back of your mind when you read the term “artificial intelligence,” in reference to the value you add to your organization? That hard knot in the back of your throat is recognition that we’re now entrenched in technology innovations that will shape our world — and jobs — forever.

Last month, in a heavily shared and discussed AdExchanger interview, former AOL Sales Chief Jim Norton said ad dollars are swinging back to content.

Norton explained that high-end premium digital content was scant, so “advertisers have quickly realized that in an algorithmically driven world wired to a race to the bottom on both scale, efficiency and pricing, you end up with this real need to get back to a premium ad environment and user experience.” (Note: He was not replaced by AI, and instead has been snapped up by Condé Nast to help the mass media company position itself as a key digital premium player.)

That algorithm is artificial intelligence, or “AI,” and you have likely read this term or heard it in meetings and conferences more than once this month. This buzz-term du jour is everywhere — at dmexco and Advertising Week, in the halls of media agencies, publishers and tech platforms.

But backed against a wall, could you explain what it means? Let’s unpack it and agree upon a definition: AI is machine learning used to predict the likelihood that a given tactic will produce a certain outcome and that the amount spent on that marketing tactic will produce desired advertising results.

Human behavior vs. machines

But the algorithms that produce marketing predictions still don’t possess an intelligence that allows them to actually mimic a human. It’s hard for machines to mimic human behavior and logic because humans can be so illogical! Who would have predicted the US GOP candidate and subsequent tone of the race? The success of Pokémon Go? That Brangelina wouldn’t last forever?

So what’s the reason behind our kind’s weird, and in some cases, clearly irrational decisions? D. Kahneman, who won a Nobel Prize for his studies, investigated why it is so hard to strictly use a rational model to predict human behavior. Kahneman discovered (and wrote a great book about) two systems which control how we behave and think — a fast, instinctive and emotional mode that is irrational and easily affected “system 1,” and the rational, logical, trainable and easy-to-predict “system 2.”

The implication of his studies in the AI fields is that no matter how many layers of automation we wrap our data in, or how accurate the algorithms become, humans still use their unpredictable “system 1” in their decisions.

This offers an opportunity to combine data science with social science, and it requires a human touch.

IBM smartly frames this as an opportunity for “augmented intelligence“: humans and machines working together. In an event addressing CMOs, the company explained how they identify a brand’s need to be addressed, then look at their tech stack and the data within their tech stack and reorganize it into something useful to meet the need of that advertiser.

It’s easy for marketers to calculate all of the different factors influencing our “system 2” by using a data-driven algorithm to get a well-based, rational targeting strategy. However, it’s harder to find those who understand “system 1” and who say, “Let’s improve on this data by adding causation to correlation.”

Norton clearly recognized the importance of the missing irrationals as he pointed out that “while there’s still a proliferation of new content being created there are fewer professional content producers. High-end, premium, curated content continues to be scarce.”

Machines process large amounts of data at scale. They do the repetitive tasks, freeing up time for you to think about the machine’s learnings and unlock value for your business.

What it means for publishers and brands

What does this mean if you’re a publisher?

Don’t fear the AI reaper. Embrace your premium abilities and the data-driven tools knocking down your office door.

You’ve invested substantial time and money in building your content teams. The publishers rising to the top of the pile are those who are staying true to their DNA while embracing new technologies and AI analytics — e.g., The Washington Post, CNN, Little Things, The Economist.

Your value is in your audiences, your writers speaking to those audiences and the technology available to create a rich brand experience for those audiences.

What does that mean for savvy media planners and brands?

Increase your programmatic education, while still embracing your natural creativity and humanity. Allow machines to do their magic at scale, and then demand rich, transparent reporting that combines campaign data and intelligent context.

Through your experimentation and investigation, strive to find simple, timeless answers, even though they are obtained through incredibly complex means.

The programmatic insurgency can sometimes feel like a sealed black box of dangerous job-ending algorithms and an endless data supply that leads to chaos or distraction. But for informed marketers and premium publishers, at the softened edges of the black box is exactly where the warmth of humanity shines and true insight is born.

Strengthening your team’s competence in the shiny new automation tools of the toolbox but keeping focused on your target/campaign goals will allow for the progress we owe to the brands and ultimately the consumers.

Ask yourself — are you driving sales? Are you building the brand? That’s the goal. AI is just the gas in the engine that drives you.

[Article on MarTech Today.]

Some opinions expressed in this article may be those of a guest author and not necessarily Marketing Land. Staff authors are listed here.


Marketing Land – Internet Marketing News, Strategies & Tips